Awesome, it is nice seeing someone who mutters to themselves like I do. People often remind me of that. I am building a computer processor entirely out of transistors made from 2N2222 transistors. It is fun for me. Keep being you no matter what, you are a gift to the maker community. Many makers end up with so many projects they never fully complete one.
Custom CPUs are always amazing to see, and doing it with transistors seems that much more challenging, which is awesome. And thank you for your kind words!
I am right now Developing a full 1970s era Supercomputer from only NAND gates (mostly because cheap) and its a really challenging task. i also have a upcomming project to make a Computer/CPU from transistors only. i am really facinated by everything logic, and its so cool to have your own computing things...
@@curiouscomputer I'll be sure to watch your videos. I thought I was replying to the original commenter saying he was building a CPU out of discrete bipolar transistors. Now THAT'S going to use a LOT of power!
Dude this is amazing! Somewhere I've got the 486 processor from my family's first computer in the early 90s. I was allowed to take the computer apart once we upgraded, and I knew that the processor was where the magic happened, although I had no idea how. But I kept the processor, even though I was like 7 years old, and I'm pretty sure it's still in my collection. As I got into robotics with the Basic Stamp 2 and started to look for more powerful controllers, I always wanted to put the 486 into a project, but it was entirely inaccessible to my 12 year old self, both intellectually and on a hardware level. It's super cool to see a breakout for this old processor! You broke one of the main barriers to accessibility. Can't wait to see what you make out of this! Instant sub.
My first CPU is AMD 486 DX2... I used it for maybe 5 years before broken. I salvaged the keyboard, Hard Drive and the CD-ROM but I didn't bother with everything else... I kinda regret it...
Wow! I did not know the 486 could be run at such low clock speeds. I'm studying to start working on something like this myself. Every 486 datasheet I've ever seen specifies the minimum clock speed as 8MHz. Being able to run it as this speed would make things a lot easier in the early debugging stages.
I wasn't sure it would work myself, but being that it's a CMOS chip it seemed like it should work that slow. It's still possible there might be other issues in the future, but it's been fairly stable in my testing so far.
@Lurch Manually running the clock only sort of works, at least for a dozen cycles or so. After that it seems like if the duty cycles are far enough out of sync it goes into shutdown mode until it gets a stable clock again.
The 486 does have an internal 8 KiB 1st level cache. I don't know if it needs to be refreshed all the time but if that is a requirement, than it would definitely limit the minimum clock speed.
This is to my knowledge the first series about making a "breadboard" 486 computer. I've searched before but i never found any. Somehow i got recomended this and im very happy. I also have a 486 that i want to make a computer out of but I plan buffering the address and data bus with octal bus transceivers(I dont want to damage the cpu).I also plan to use an arduino or something else for clock,reset and other misc stuff its more easy and convenient(like have the arduino be controlled by serial)
Fascinating project! I've seen my fair share of vintage motherboard repair videos, but it's not often I see someone experimenting with a 486 like this, let alone designing a breakout board to do so! Wishing you the best on this project, can't wait to follow along.
Awesome! An idea came to my mind: adding status LEDs to the buses directly on the board would look great, but also could could double as a test bench for the CPU, for diagnosing ones taken from retrocomputers.
I've played around with that, and now that I'm more comfortable with SMDs it would be easier. There is a limited amount of space as the board is 100x100mm and 4 layers, anything bigger then that has a huge jump in price.
Wow! I am absolutely living for this! I’ve long wondered what kind of homebrew type stuff could be done with the venerable i486. Can’t wait to see more! Gonna be absolutely glued to this channel!!
I’m not sure they exist for 486, but for the older 386 and 286, there were several CMOS "static core" variants produced for embedded purposes, allowing you to run the clock at whatever rate you wanted, even stopping the clock completely without losing state. Some examples are Intel’s 386EX and the Harris CS80C286. Many CPUs of the era used dynamic memory for internal registers to save die space, and slowing the clock too far results in them not being refreshed frequently enough. It may be able to run very slow for now, but if you start trying to use the internal caches, the MMU, etc., fun errors might start happening :). Assuming you weren’t already alluding to it with your mention of the debounced switches, you can always go for the "button that asserts ready for a single bus cycle" trick. Edit: supposedly the Cyrix Cx486DX was a static core design.
Compared to the clean-cut high production value I'm used to on RU-vid, the raw footage here shocked me with how much it felt like I was watching myself Quickest 20 minutes on RU-vid!
I remember back like 2004/5 in high school some friends and I were all in agreement that like, 386s and 486s are far too complicated for a homebuilder to ever build their own computer around. Granted that was in the days before easily designable and quickly available custom multilayer boards. But it's an assumption I'd been second guessing lately, so seeing this is extremely cool. Very nice work.
At this point I believe the Pentium / Socket 7 systems are the "very difficult" point. The bus protocol was never officially public, although the AMD K6-2 manuals have a very good description of the bus timing. It’s a synchronous bus so timing becomes a difficult problem. Totally solvable using FPGAs, but the signal count is high (like 250), so you’ll have to use a BGA variant.
Thank you, I'm glad you enjoyed it! The format is inspired by a few other creators I follow, especially @weirdboyjim . I will probably be having some edits for timelapse in the future, or to cover problems with my current setup, but not to remove any content. Seeing the mistakes are half the fun for this sort of project.
@@Fox-Tech Time lapse for certain parts is not a bad idea :) I was curius about doing something with an x86 but when I read that the minimum frequency was in the MHz range i hesitated. Seeing that you can i fact run them in the KHz range inspire me to build something. Can't wait for you next episode in the 486DX series :D
"The standard NOP... So yeah we're just going to start telling it do do nothing and if everything works it should do nothing. but it's the nothing that we are telling it to do. which is better than the nothing we're not telling it to do. That made more sense in my head." - Dude, that is hilarious. I want it on a T-Shirt just to make the 99.9999 % of the world say , "WTF does that mean?" It only makes in the heads of those who are fascinated by this kind of stuff. A very small club indead (I am a member of said club ) Carry on !! Cheers
This reminds me of an idea I had for repurposing older but functional CPUs: turning them effectively into some sort of microcontroller board, like an Arduino. Obviously they would be a bit bigger BUT they could be standardized.
"A full breadboard computer with a 486." Ambitious! Nice. :) It's really interesting to see how the 486 connects up to everything. Nice to know it can run at much lower clock speeds than advertised.
This is a very cool project. However, I strongly recommend that you seriously consider increasing the number of ground connections. You don't appear to have any on to the board apart from the power connection. With wide busses, it will become very important that the ground for the signals doesn't bounce around when the signals switch. Even if you are only clocking at (say) 100kHz, the signals still transition between high and low logic states at the same speed as if you are running at full speed. The rapid simultaneous charge or discharge of multiple signal lines causes noise spikes in the local signal ground, which affects how other devices see the signals. (Research ground bounce for more info.) I'd recommend one or two grounds for every eight data or address lines, and more grounds where you are grouping control signals. While things may work OK reading static instructions from switches, it will very quickly get hard to debug unless you can keep the signals free of too much noise. This will only get harder with a breadboard compared to a good PCB design.
Thank you. That's very good to know, I will research that and try to improve things. Right now the board does have a layer that's just ground plane, I don't know if that will help any in this case. Still, I will be looking into that and trying to apply it on future PCBs.
@@Fox-Tech Certainly a ground plane is a good idea, but you also need to minimise the ground impedances between CPU and each area of the design that it's connected to such as the data buffers that you added, clock sources, decoders, anything driven by the address lines, and so on. Ideally, as well as the ground from the power supply, you will have direct ground paths from CPU to each other section, and they will also have their own ground connections back to the power supply as well. One of the difficulties is that the ground voltage spikes caused by fast ground current transients can be extremely short and very localised making them hard to measure. (Q... when is zero ohms not zero? A... When it's a 50 nano-Henries of inductance and the current spikes from 2mA to 100mA and back in less than 10 nano-seconds.) This all leads to the issue that each chip will see the same signal differently, based on it's local ground (at the chip inside the package) and that will be somewhere that you cannot measure even with a decent oscilloscope. This is why it's better to design high speed digital circuits with an over abundance of caution regarding signal grounds, rather than trying to fix the problem afterwards. But don't be discouraged. With the right research (which you've already demonstrated in terms of how the 80486 talks to the data-bus, for example, this could become a very interesting and rewarding project. I am watching with interest and admiration. (I have designed plenty of fast complex logic, PCBs, etc, but never around a 486 - and I have a DX33 in my junk box somewhere...)
Thanks for making this video. I'm glad it's not just me who mutters along to myself while working. You can always cut/ speed up sections of putting wires in if it's more comfortable. Great level of detail as well. It's easy to go too simple or too complex
Really nice to see people push the limits of breadboard computers beyond typical 8-bit DIP CPUs. This project looks amazing and made me an instant subscriber. Hopefully, in the not-too-distant future, this will inspire someone to go even further and make a breadboard computer using a contemporary CPU. Even a brand new Celeron would be incredible to see. Still, your work with the 486 is amazing and deserves far more attention.
I really Love having found you, you are extremely smart, and even though the video is kinda slow pace for todays youtube standarts it was extremely satisfying to watch and i really enjoyed it a lot. id love to see more of your content
This is really cool! As someone with a passing interest in this type of old computer some of this flies over my head but it looks like a fun concept. One thing I would suggest is maybe adding a socket to the breakout board so you could use other 486 chips from like AMD or CyrIx
There is a socket under the chip, but it's a standard pin socket and not a ZIF socket (which are much harder to find, might see if I can scavenge them from old motherboards in the future) so it's not really visible.
Love the concept of this. Watching to see how far you can go with off the shelf modern parts. Could be a way to have a nice PCI setup as the 486 PCI boards are either really bad or extremely rare and expensive.
PCI is way more involved than what's needed for a bit of RAM and ROM. My EE course where we did microprocessor interfacing we just did PC bus (as in original 8086->style) which is simple enough to do with simple gates and latches. For PCI you need a fairly complicated state machine, especially for the discovery and initializing.
Really nice Project, looking forward to the next episodes :) First time I got the idea of a the 486 as a managable processor and how it operates, havent looked into it very much before, but inspires to learn more about it. Keep it up, especially with the awesome one take :D
I like it! Just today i ordered a PCB what extract the pinout of a 186 to a ISA Bus (for sure with custom pinout, but useable on a standard backplane). I want to do the same "Selfmade Computer" with it. The 186 uses A0 and BHE to seperate the lower and higher Byte of the 16bit Data. And yes, you can mirror/overlap a single memory Chip/adress space to two different locations 😅 Its nice to know that the 486 can work with 8bit components. And one of my first tests will be the lowest working frequency because i want to see blinking LEDs too
That sounds like fun. I don't think the 186 gets used enough, so glad to see someone doing something with it! In 16-bit bus mode, the 486 has the same A0/BLE BHE signals. That's one of thing that seems consistent throughout all revs of the CPU.
@@Fox-Tech There are some DIY with the 80186/88 in the WW - but just a very few. Most of them are SBCs copied from very old schematics or pc-magazines. I want to start from scratch with something like a computer and not PC/Amiga or somthing compatible.
Oh wow, this is neat! I had no idea you could interact with an intel CPU like this without having a full PCI bus, north/south bridge, memory channels etc. It will be super interesting to see how far you’re able to get, eg if you write your own simple BIOS etc. Nice work man 👏
It's funny that although tech has evolved (boomed) so much in the last 100 years or so, and there are things that are way faster and more efficient than this aging 486, yet I'm still finding it insanely cool to look at old tech, especially when a "modern twist" is added to them. Very very cool board, now please add a screen and run Doom on it.
I am interested to see where this goes (and yea its similar to most processors, but its still interesting as most people wont go past dip packaged chips)
I could be mistaken, but as I gather, the 486 was pretty much the last of the "similar to other processors" before things like low voltage levels and specialized busses became a thing. The beginning of that trend is touched-on, here, in that even though you can tell the 486 to access individual bytes, it expects them on different data-bits depending on their alignment. Not difficult to do with TTL, but not "easy", either. (imagine its being connected to an 8bit ISA card!). So, at some point it became almost essential to have bridge chips to interface with the outside world, even if that's nothing more than a boot ROM. So, in a lot of ways the 486 is maybe amongst the last of the "similar" CPUs (Z80, 6800, even 8088) (and comparatively *really powerful*) that folk like us *can* "breadboard". That was a very exciting finding...
Meh, people say the wrong word all the time. Half the RU-vidrs I watch don't even correct themselves in post. Don't beat yourself up about it. Fun video.
I used to design Intel based systems i386DX ,i486,i960 ,i860 .All had similar buses and needed epld 10 ns ( very fast back then ) to latch some of the bus control signal .The code was given by Intel in the hardware manual .
Oh, that's cool. Feel free to tell me when I'm doing something completely wrong then! Actually, I think I ran into some of that code in question today, while looking for something else.
This is truly FUN 😀 My bet you have started a whole new trend. Now to integrate it into the Arduino world - then you will suddenly get support as you never dreamed of ... I just sub'ed as I find this project very interesting indeed !
I'm looking forward to the next episodes of this :) Also: I'm wrecking my brain on how to connect a single 8bit RAM to all four parts of the data bus. I guess I'd aim at 4 switchable bus tranceivers (74xx245?) connected to the byte enable lines and a 4xOR from all byte enables to the /CS of the RAM...
The proper way to introduce the 486 is "The DooM CPU." BRB, going to go invent high-speed solderless prototyping, so this guy can build this thing and go full speed. To play DooM of course.
i'd been curious to try something like this myself if i had some old processors to mess with. Would be fun to turn one of those old cpus into an impractical microcontroller or something.
I really like the idea and your presentation! A friend gifted me a beautiful 33MHz 486 a while ago, but I haven't had a way to use it yet. I wanted to make a board like this, but the datasheet seemed a bit too complex for a weekend project. Are you planning to share the board design? I'd even like to use this in my own YM2612 retro synthesizer design.
I'm glad you like it. I wasn't sure of exactly how to handle this, and tried a few different things before settling on this one. I do have an slightly older rev of the board on my github, hope to get it updated soon. github.com/FoxCutter/Tools/tree/master/486%20Breakout I should also add some more documentation on how it should be used with a DX2 and possibly DX4 board.
The split into separate code address space and data address space was also available on the Motorola 68000 series as well, I think starting with the 68020. I hadn't heard of any 68K systems that actually used that feature either. In theory it could allow for better code security by locking out modifications to running code. It might also make it easier to map in ROMs without having to carve out a block of the address space.
About as far as I go with BB based computers is MC68k based beyond that when I go with higher pin count PGA CPU’s I move to wire wrap or PCB’s which is MUCH cheaper than wire wrap & can get good signal integrity.
I never thought it would be possible to do this with an Intel processor. It always seemed complicated to me compared to mc68k and simpler processors... fascinating.
That breadboard power supply cannot really handle the current requirements of a 486dx which are about 500ma and peaking at 700ma. The voltage regulator (a1117), which while rated at 1A, will get really hot and shatter or go into thermal shutdown depending on the voltage dropped -- it would be much safer to get a better power supply like a old PC power supply or a switching brick rated 5v at a few amps 2+
That's good to know, thank you! I'll keep an eye on it and look into finding something better. Not sure what the best solution would be, maybe an atx breakout board so I can hook up a mini-atx power supply (not sure a pico would be enough amps). I was already considering moving to atx when I made the back plane, so this just helps with that.
I'm probably going to go with FORTH instead of basic, as I've already done a full implementation of it for the z80, so I have a lot of stuff already done. Not quite as old school as Basic though.
Hey this is cool! Back in 05 I designed and build a computer using an 8085. I always thought it would be cool to make one based off the 68020, but that's a lot of wire wrapping! I'm subscribing to follow along!
Honestly, ordering custom PCBs is so cheap these days that I have become lazy and go for a PCB first … problem is what to do with the other PCBs I didn’t use because most of my projects are one-offs and you get a minimum of 5…
@@teknoman117 I would do the same. It's just that every time I sit down with the software It feels like there is a large learning curve. I got enough going on without having to take the time to learn it.
But where's the fun in that? In theory I could just replace the bios layer and run dos on whatever I end up with. I don't really know what the end goal is other then finding out.
@@Fox-Tech The fun is to make an IBM compatible PC with your own designed PCB motherboard. And in the end, the reward is everything to have a PC that runs everything that ran on a PC back then. Nothing else can compete with it.
pretty cool.. but i personally would break it out as a standard PC-104 board.... the reason being there are plenty of peripherals and other stuff that is designed perfectly for something like this. in fact you can still buy (if you're lucky) pc104 386 and 486 and lower end pentium boards.
I certainly didn't know that the 486 only uses 30 address lines in 32bit mode. But it makes sense, since the Program counter probably does not go beyond FFFF FFFF.
It's on my github github.com/FoxCutter/Tools/tree/master/486%20Breakout I did update it the most recent version (though a V3 looks likely), and added some notes to it. I will probably move it, along with everything else, to it's own Repo in the near future.
I have a 486sx waiting for something just like this. If you are working on making an 8bit bus, it might be worth looking at the z50 or rc2014 buses. The z50 is only a 16bit address bus, but you could still play in the bottom 64k. The extended rc2014 bus has 16bit data and 24bit address, but could be possible to do 8bit/16bit data & 16/24bit addr.
Depending on the revision of the SX, the breakout board could work for you. The main issue is the NMI pin was put into a different location (A15) which is used for a FPU related function on the DX, this was changed in the SX2 to put it in the same location (B15) as the DX. As for the bus, my hope is to eventually have a 32-bit RAM bus with an 8-Bit ROM, but I will look into the ones you suggested as well.
I don't know if this will help, but Intel made a MUCH smaller version of this CPU in the late 90s. They were used in embedded applications at the time. There was one in my old Lucent Portmaster that was about 1/4 inch on a side. Maybe it's worth hunting one down?
Thanks for your great sharing. Should we worry about the cooling problem? If not, I'm curious if it is caused by such a low clock frequency or by 486 itself.
Later models of the 486 did need a heat sink, but this one can run without it. And as it, even running at 4mhz for a half an hour or so still leaves the chip itself cool to the touch. That could also be because the pcb has two copper layers which might be helping with some of the heat dissipation. That might change if I ever get it up to the full 25mhz that the chip is capable of, but for the moment it appears just the ceramic provides enough passing cooling.
Intstantly subscribed when saw video in my recommendations. But please do not use that breadboard power supply for power regulation, regulators can fail easily and feed the processor input voltage directly.
This is so cool!! now, one question for this is that this is a somewhat modern processor for most breadboards (most use a 6502 or maybe 8086) and this does run x64. could you have a breadboard Linux computer? the last time a 486 cpu was supported on linux was four months ago! and maybe in theory you could employ enough breadboards and chips to replicate the hardware of a motherboard so you could put in a video card! besides. a breadboard can do anything a pcb can do but only bigger! it would be so cool seeing that happen! but I feel like it would be unlikely as it could take too much time.
and yes, linux did actually support this for that long. long enough for this to possibly run on a breadboard. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-M4EPXwUV3aQ.html
